Investigation of the influence of geomechanical and hydrogeological properties on surface uplift at In Salah

Investigation of the influence of geomechanical and hydrogeological properties on surface uplift at In Salah

Journal of Petroleum Science and Engineering xx (xxxx) xxxx–xxxx Contents lists available at ScienceDirect Journal of Petroleum Science and Engineer...

1MB Sizes 0 Downloads 18 Views

Journal of Petroleum Science and Engineering xx (xxxx) xxxx–xxxx

Contents lists available at ScienceDirect

Journal of Petroleum Science and Engineering journal homepage: www.elsevier.com/locate/petrol

Investigation of the influence of geomechanical and hydrogeological properties on surface uplift at In Salah ⁎

P. Newella, , H. Yoonb, M.J. Martineza, J.E. Bishopa, S.L. Bryantc,1 a b c

Sandia National Laboratories, Engineering Sciences Center, 1515 Eubank SE, MS0840, Albuquerque, NM 87185, USA Sandia National Laboratories, Geoscience Research and Applications, Albuquerque, NM 87185, USA The University of Texas at Austin, Petroleum and Geosystems Engineering, Austin, TX 78712, USA

A R T I C L E I N F O

A BS T RAC T

Keywords: Geomechanics Geological CO2 sequestration In Salah InSAR Inverse modeling Modeling Mulitphysics

Coupled reservoir and geomechanical simulations are significantly important to understand the long-term behavior of geologic carbon storage (GCS) systems. In this study, we performed coupled fluid flow and geomechanical modeling of CO2 storage using available field data to (1) validate our existing numerical model and (2) perform parameter estimation via inverse modeling to identify the impact of key geomechanical (Young's modulus and Biot's coefficient) and hydrogeological (permeability and anisotropy ratio) properties on surface uplift and the pore pressure buildup at In Salah in Algeria. Two sets of surface uplift data featuring low and high uplifts above two injection wells and the maximum change in the pore pressure due to CO2 injection were used to constrain the inverse model. Forward simulation results with representative parameter values from the literature match both low and high surface uplifts reasonably well and predicted the maximum change in the pore pressure. In particular, forward modeling results with estimated Biot's coefficients for reservoir and caprock layers, match the observed uplift well, highlighting the significance of Biot's coefficient in coupled reservoir and geomechanical models. Parameter estimation with 12 parameter sets for both low and high uplift data demonstrates that multiple sets of parameters can match the observed data equally well and the inclusion of the pore pressure data is critically important to constrain the parameter solution during inverse modeling. For a majority of cases, estimation results for both low and high uplift data show the vertical intrinsic permeability and Young's modulus of the reservoir remained close to 13 mD (1.3×10−14 m2) and 10 GPa, respectively, suggesting that these parameters may represent the actual effective properties. Additionally, higher correlations between reservoir permeability and caprock's Biot's coefficient with high surface uplift data were observed consistently under the pore pressure constraint, suggesting the inclusion of the pore pressure constraint is required to estimate the proper values of coupled flow and geomechanical properties associated with different surface uplift data. Overall, this study suggests that given limited data, including Biot's coefficient, in addition to permeability and Young's modulus can enhance parameter estimation of the geomechanical response during GCS.

1. Introduction Over the last decade, geologic carbon storage (GCS) has been proposed as a promising technology to reduce CO2 emission to the atmosphere. It is critical to understand geomechanical processes and impacts from CO2 injection to ensure that CO2 can be securely stored over geological time. Coupled multiphase flow and geomechanical models can be used to understand and assess effects of increased reservoir pressure by CO2 injection on the geomechanical response. One of the prominent field demonstration projects is the In Salah Gas project, located in Algeria, where CO2 recovered in natural gas ⁎

1

production was reinjected into a sandstone reservoir formation (Eiken et al., 2011). The spatial distribution of surface uplift was successfully evaluated using data obtained from the satellite based inferometry (InSAR) (Vasco and Novali, 2008; Vasco et al., 2008). This uplift data and other geophysical data have been used to investigate reservoir properties and coupled flow and mechanical processes (Rutqvist et al., 2009, 2010; Preisig and Prevost, 2011; Shi et al., 2013). In the past, predictions of the impact of CO2 injection on the geomechanical response, such as surface uplift, have been performed with relatively simple models. The use of simple models has been

Corresponding author. E-mail address: [email protected] (P. Newell). The University of Calgary, Chemical and Petroleum Engineering, Calgary, Alberta, Canada.

http://dx.doi.org/10.1016/j.petrol.2016.11.012 Received 9 October 2015; Received in revised form 8 September 2016; Accepted 11 November 2016 Available online xxxx 0920-4105/ © 2016 Published by Elsevier B.V.

Please cite this article as: Newell, P., Journal of Petroleum Science and Engineering (2016), http://dx.doi.org/10.1016/j.petrol.2016.11.012

Journal of Petroleum Science and Engineering xx (xxxx) xxxx–xxxx

P. Newell et al.

justified because of nonlinearities of coupled multiphase flow and geomechanical response, the associated computational expense, and/or a lack of field data. Rutqvist et al. (2009, 2010) investigated the relationship between the surface uplift and pore pressure change as well as deformation within the injection zone at In Salah, using coupled reservoir and geomechanical modeling. They showed consistency between the simulation results and the measured data from InSAR on the surface uplift. They demonstrated that volumetric expansion of reservoir rocks and surrounding shaly sands may contribute to the surface uplift. This volumetric expansion depends on both permeability and elastic properties of the reservoir and overlying caprock. Preisig and Prevost (2011) presented a two-dimensional (2D) fully coupled multiphase thermo-poromechanical model for simulating CO2 injection at In Salah. Their 2D-model over-predicted the surface uplift. They concluded the lack of the accurate field data to be the reason. They also demonstrated that creation or reopening of fractures can be attributed to temperature differences between injected fluid and reservoir rock. Reservoir characterization of faults and fractures has been performed to enhance the understanding of CO2 flow in fractured rocks at In Salah (Iding and Ringrose, 2010; Pamukcu et al., 2011; Deflandre et al., 2011). Iding and Ringrose (2009, 2010) confirmed the presence of fractures and small faults in both the reservoir and the lower caprock, based on the long-term performance data of In Salah field. They concluded that despite the clear evidence of fractures in the reservoir, a thick caprock layer provides an effective hydrological and mechanical barrier to CO2 leakage. Recently, Smith et al. (2011, 2013) also investigated possible fracturing within the reservoir and lower caprock near one of the CO2 injection wells at In Salah. They concluded that at the given injection rates, induced fracturing into the upper caprock should not occur and the possibility of CO2 leakage through induced fractures is low. Geophysical inverse techniques using InSAR and seismic data have been employed to estimate reservoir volume change and fault-fracture aperture change after CO2 injection (Vasco et al., 2010). The results indicate that CO2 associated flow can extend up to several kilometers through the fracture/fault zone in the reservoir formation at In Salah. However, all of these models did not fully employ coupled multiphase flow and geomechanical modeling. Accurate identification of geomaterial properties is essential for predicting fluid flow and geomechanical response. In reservoir engineering literature, the process of calibrating model parameters with dynamical data is known as “history matching”. Using a multiphase flow simulation, Pamukcu et al. (2011) manually performed history matching at the In Salah site to calibrate the porosity and permeability of matrix and fracture network in the reservoir using bottom hole pressures, CO2 injection rates, and a CO2 breakthrough time at a monitoring well. Although they matched observed data reasonably well, they concluded that coupled multiphase flow and geomechanical modeling is required to confirm their simulation results. Recently, Shi et al. (2012) performed history matching with the temporal changes in the maximum vertical uplift to estimate Young's modulus. The reservoir model with stochastically generated porosity and permeability static fields was first calibrated with observed dynamic bottom hole pressures to estimate fracture transimissibility of reservoir and lower caprock layers independently. The calibrated reservoir model was then imported into a coupled reservoir-geomechanical model in order to calibrate Young's modulus of the lower caprock with surface uplift data near one of three injection wells. The calibrated model was then used to match InSAR surface uplift observed at two other injection wells, demonstrating that model predictions capture the overall trend well, but the mismatch was much greater. Despite their history matching with manual tuning of 1–2 parameters, the physical models were built upon the best estimation of reservoir and geomechanical characterization. Their results clearly highlighted the importance of coupled reservoir-geomechancal modeling to evaluate the performance of CO2 injection at the reservoir scale. Zhou and Burbey (2014) mentioned despite the fact of limited hydrogeological and geomechanical informa-

tion about the host rock formations, development of these sophisticated surface monitoring techniques such as InSAR and GPS, can yield critical information about the rock formations, which can be used in numerical modeling to monitor the fate and transport of the injected fluid. In a more general aspect of reservoir simulation, Nanayakkara and Wong (2009) provide an interesting discussion on analytical and numerical modeling of surface uplift due to subsurface injection. They investigated multiple cases and showed the importance of the boundary, in terms of both the location of the boundary and the boundary conditions. Their study revealed that more realistic results could be obtained through fixed displacement boundary conditions at the bottom boundary as well as selecting sufficient lateral extent. Khakim et al. (2012) also tested a two-step inversion method to estimate the distribution of reservoir deformation and volume change due to steam injection using a 3D synthetic problem. The depth of injection point was first estimated with approximate modeling of the deformation, followed by the accurate estimation of the reservoir deformation and volume change due to the surface uplift with the InSAR-derived surface deformations. However, it was not tested in a multiple layered system whose mechanical and hydrogeological properties can vary significantly. Aoyagia et al. (2013) have developed a numerical simulator for environmental impact assessment. The In Salah CO2 storage case was used to validate the simulation. They also performed a sensitivity analysis on some parameters, such as caprock permeability, porosity and Young's modulus of the reservoir. They found that caprock permeability can significantly affect the surface uplift. Over the past decade, optimization, sensitivity, and uncertainty quantification of multiphase flow models during GCS have been developed in the literature (Espinet and Shoemaker, 2013; Wainwright et al., 2013; Tavakoli et al., 2013). However, automatic parameter estimation of coupled multiphase flow and geomechanical models has not been thoroughly investigated. A study of geomechanical deformation due to CO2 injection by Verdon et al. (2013) highlights the importance of systematic geomechanical evaluation prior to CO2 injection. Because of the non-uniqueness of the inverse problem and model uncertainty, multiple parameter sets with various starting points need to be compared. This allows evaluating the sensitivity of correlated parameters when calibrating the model (McKenna and Pike, 2013; Yoon et al., 2013). The aim of this study is to identify and rank the importance of key geomechanical and hydrogeological parameters using coupled flow and geomechanical simulations to have a better understanding of the surface uplift at In Salah, Algeria. Because of the lack of field data, there is a high uncertainty involved in the model parameters. This uncertainty arises from geomechanical, hydrogeological properties and the coupling parameters. In this study, the sensitivity of the surface uplift to some of these critical parameters will be investigated. Specifically, two sets of surface uplift data featuring low and high uplift above two CO2 injection wells are used. In addition, the maximum change of pore pressure due to CO2 injection is included to evaluate the impact of pore pressure constraint on surface uplift during parameter estimation. After validating our coupled forward model with the representative parameter values in the literature, simulation results are used to evaluate the significance of permeability, permeability anisotropy ratio, Young's modulus, and Biot's coefficient on surface uplift and pore pressure increase. Parameter estimation with 12 different sets of parameters is performed for both KB501 and KB503. Estimated sets of parameters and resulting pore pressure calculations are used to evaluate the significance of parameterization and the inclusion of the pore pressure constraint on the geomechanical response. It should be noted that the focus is on KB501 and KB503 and all of the simulations are based on these two injectors. Due to the complex nature of KB502 and many unknown factors involved in its surface uplift, this well will not be discussed in this study. 2

Journal of Petroleum Science and Engineering xx (xxxx) xxxx–xxxx

P. Newell et al.

Fig. 1. Krechba gas field (Rutqvist et al., 2010). The red squares show the simulation domains for KB501 and KB503. The dashed domains show the quarter symmetry sections actually used in the simulations.

sandstone overburden aquifer, and (4) 2180 m-thick Cretaceous sandstone base aquifer (see Fig. 2). Supercritical CO2 was injected via a 1000 m-long horizontal well. For simplicity, it was assumed the model domain and all conditions are identical for both the KB501 and KB503 well regions.

To the best knowledge of the authors, this is the only study investigating the influence of these properties on the surface uplift through multiple, both single and combined parameter estimations at In Salah. Moreover, the effect of Biot's coefficient and permeability anisotropy ratio on the surface uplift at In Salah (KB501 and KB503) has not been explored by any other researchers. Pore pressure constraint is another unique aspect of this article investigated in terms of surface uplift in both KB501 and KB503. It should be noted that a complex numerical model was not the objective of this study, therefore, the model was simple, yet still sufficient to capture the physics of the problem. For instance, this study did not account for heterogeneity in rock layers.

2.2. Coupled flow and geomechanical model The Sierra simulation software developed at Sandia National Laboratories was used for the numerical analysis. Sierra is an engineering mechanics simulation code that includes a suite of highly parallelized finite element analysis modules for different physics (Notz et al., 2007; Martinez et al., 2013). Martinez et al. (2013, 2011) provided the details of the numerical modeling of CO2 injection in GCS systems. Here, we briefly describe the key aspects of coupled multiphase flow and geomechanical model. In this work, the Sierra Arpeggio module, which is a coupling module between the Aria module for multiphase flow and the Adagio module for the solid mechanics, was used. The Aria flow module uses the mass balance of the two-phase system described as

2. Model description 2.1. Model domain As shown in Fig. 1, the simulation domain of a given injection well assumes quarter symmetry. Each domain is discretized into a 3Dmodel of size 5×5×4 km3. Following the literature (Iding and Ringrose, 2010; Raikes et al., 2008), the four distinct layers are: (1) 20 m-thick mudstone reservoir, (2) 900 m-thick Carboniferous mudstone and tight sandstone caprock layer, (3) overlying 900 m thickness of Cretaceous

Fig. 2. (a) Details of the stratigraphic layers used in the model. (b) Schematic of the 3-D simulation domain and finite element mesh.

3

Journal of Petroleum Science and Engineering xx (xxxx) xxxx–xxxx

P. Newell et al.

⎞ ⎛ k ∂(ρw ϕSw ) ∂(ρn ϕSn ) = ∇·⎜ρw rw k ·(∇pw − ρw g ) ⎟ + Qw ∂t ∂t ⎠ ⎝ μw ⎛ krn ⎞ = ∇·⎜ρn k ·(∇pc + ∇pw − ρn g ) ⎟ + Qn , ⎝ μn ⎠

surface is fixed while the top surface is free to move. Iding and Ringrose (2010) reported that the storage unit has an initial temperature of approximately 90 °C. Thus, for the simplicity, the system was assumed isothermal at 90 °C. The average field injection rate was 0.2 megatonne/yr (Mt/yr) for both KB501 and KB503 Cavanagh and Ringrose (2010). It should be noted that the uplift data in this work were obtained graphically from Rutqvist et al. (2010).

(1)

where w and n represented the wetting phase (water) and non-wetting phase (CO2), respectively. Additionally, ρ is density, ϕ is porosity, S is saturation, kri is relative permeability tensor of phase i, μ is viscosity, k is intrinsic permeability tensor, pc is capillary pressure defined as pc = pn − pw , g is the gravity vector, and Q is the flux term of phase i. The pore space is assumed to be fluid saturated Sw + Sn = 1. The Adagio, solid mechanics module solves the balance of linear momentum,

2.3. Parameter variation Single parameter studies were preformed to observe how the value of each parameter influences the surface uplift. Results of the single parameter study were used to setup initial values for automatic inverse modeling, which will be described in the next section. The measured surface uplift at two injection wells KB501 and KB503 were used to compare simulation results with different parameter values. Since the surface uplift at KB501 is lower than that at KB503, different caprock permeability values were used for KB501 and KB503 (Table 1) based on Rutqvist's suggestion (Rutqvist et al., 2010). The key parameters investigated include: caprock permeability (kz), Young's modulus of reservoir and caprock (E res and Ecap ), Biot's coefficient of reservoir and caprock (bres and bcap ), and anisotropy ratio of the reservoir permeability (α res ) defined as the ratio of the horizontal intrinsic permeability (kx=ky) and the vertical intrinsic permeability (kz).

(2)

∇·σ + ρg = 0,

where σ is the Cauchy stress tensor and ρ is bulk density of the mixture. The inertial effects were assumed negligable. The solid skeleton is assumed to behave as a linear elastic, (3)

σ = Cϵ ,

where C is the fourth-order stiffness tensor and ϵ is the infinitesimal strain tensor defined as:

ϵ=

1 [∇u + (∇u )T ], 2

(4)

where u is the displacement vector and T represents a transpose. The reservoir was initially water-filled and after CO2 injection, a multiphase flow system must be considered (Vasco and Novali, 2008). For the continuum multiphase flow, supercritical CO2 and water are treated as two immiscible, compressible phases. For capillary pressuresaturation relationship, the van-Genuchten model (Genuchten, 1980) was used. For the wetting and non-wetting relative permeabilities the van-Genuchten and cubic functions were employed, respectively. All multiphase flow parameters are identical to those used by Rutqvist et al. (2010). To account for the pore pressure effect on the stress tensor, the poroelasticity theory based on the effective stress was used

σ eff = σ − bIp ,

α res =

kh . kv

(6)

Caprock permeability for KB501 and KB503 is set to 1.0 × 10−21 m2 (Sim-KB501) and 1.0 × 10−19 m2 (Sim-KB503), respectively. Rutqvist et al. (2010) suggested two different caprock permeability values to capture the surface uplift of KB501 and KB503. In order to investigate the possibility of changing in other parameters, rather than just permeability, to capture these surface uplifts, three different cases were studied. In Sim-KB501-Biot, different values of Biot's coefficient (bres =0.55 and bcap =0.75) were used. In Sim-KB501-AnisoPerm, kz=1.927×10−14 m2 and α res was set to 0.536. In Sim-KB501-BiotAnisoPerm, different values of Biot's coefficients (bres =0.55 and bcap =0.75), plus anisotropic permeability field for reservoir were used (kz=1.3×10−14 m2 and α res =0.77). In all these cases, the other parameters remained the same as KB503, and the goal was predicting the surface uplift of KB501. It should be noted that the maximum change in the pore pressure with respect to the initial pore pressure within the reservoir, due to CO2 injection (ΔPmax ), was estimated to be approximately 10 MPa (Rutqvist et al., 2010). This maximum change in the pore pressure can be supported by bottom hole pressures measured at injection wells

(5)

where σ eff is the effective stress tensor, b is Biot's coefficient, I is the identity tensor, and p is the pore fluid pressure. Key hydrogeological and geomechanical properties used in this work are listed in Table 1. Referring to Fig. 2, the two adjacent vertical planes, the x-z at y=0 and the y-z at x=0 are no-flow boundaries, while their opposite planes at x=5 km and y=5 km are under constant pressure corresponding to initial hydrostatic condition. For the solid skeleton, initial lithostatic stress conditions are applied. All vertical sides of the domain are fixed against normal motion. The bottom Table 1 Hydrogeological and geomechanical properties of each layer identified in Fig. 2. Property

Overburden

Caprock

Reservoir

Base

Units

Vertical intrinsic

1.0 × 10−17

1.0 × 10−21(KB501)

1.3 × 10−14

1.0 × 10−19

m2

1 6 1 0.2 0.17 2100

1 20 1 0.15 0.01 2100

0.3 0.05 19.9 0.457

0.3 0.05 19.9 0.457

permeability (kz)Rutqvist et al. (2010)

10−19 (KB503)

Anisotropy ratio (α) Young's modulus (E) Biot's coefficient (b) Poisson's ratio (ν) Initial porosity (ϕ) Solid density (ρ)

1 1.5 1 0.2 0.1 2100

1.0 × 1 20 1 0.15 0.01 2100

Residual liquid saturation Residual gas (CO2) saturation van-Genuchten, P0 van-Genuchten, m Wetting phase (water) viscosity

0.3 0.05 19.9 0.457

0.3 0.05 19.9 0.457

0.315 × 10−3

0.315 × 10−3

0.315 × 10−3

0.315 × 10−3

Pa-s

Non-wetting phase (CO2) viscosity

5.6 × 10−5

5.6 × 10−5

5.6 × 10−5

5.6 × 10−5

Pa-s

4

GPa

kg m3

(kPa)

Journal of Petroleum Science and Engineering xx (xxxx) xxxx–xxxx

P. Newell et al.

optimum, which is strongly influenced by the starting value of the estimation (Yoon and McKenna, 2012; Yoon et al., 2013). Hence, multiple sets of parameters may give the same goodness of fit to the calibration data unless the starting value is close to the best solution, which is typically unknown for real complex problems. To overcome this issue, a multiple starting approach was evaluated by changing the initial values of estimated parameters. The initial values were selected from a range of possible values for In Salah.

Table 2 Five-parameter groups used for model calibration. Group

Calibration parameters

Group I Group II

Permeability parameters for reservoir only (kz,res , α res ) Permeability and mechanical parameters for reservoir only (kz,res , α res , bres , Eres

Group III

Permeability and mechanical parameters for caprock only (kz,cap ,

αcap , bcap , Ecap ) Group IV

Group V

3. Results and discussions

Permeability parameters for both reservoir and caprock (kz,res , kz,cap , α res , αcap )

3.1. Model validation

Permeability and mechanical parameters for both reservoir and caprock (kz,res , kz,cap , α res , αcap , bcap , Ecap )

The base parameters in Table 1 were used in the simulation of In Salah. The simulation results were used to validate our coupled geomechanical, multiphase fluid model. Sim-KB503 used caprock permeability of 1.0×10–19 m2 and Sim-KB501 used 1.0×10–21 m2, as suggested by Rutqvist et al. (2010). These simulations are coupled multi-physics in nature, and it is not straightforward to decouple the impact of the geomechanical and hydrogeological properties on the surface uplift. However, from Equation (5), it is clear that the distribution of the pore pressure has direct impact on the effective stress field, which influences the strain field following the constitutive laws of each layers. Fig. 3 shows how pore pressure affected the vertical total stress field and created the vertical effective stress field based on the Terzaghi's equation. Fig. 3d shows how pore pressure affected the CO2 saturation field. As a result, the displacement field was affected in the system (Fig. 3e). As can be seen in Fig. 3e, the maximum vertical displacement occurs within the reservoir, where the maximum pore pressure change (ΔP ) occurs (Fig. 3b). Moving towards upper layers, the change in the pore pressure decreases and thus the displacement field decreases as well, but most of the surface uplift is due to the injection reservoir expansion. Other works (Aoyagia et al., 2013; Shi et al., 2013) have suggested Young's modulus of 10 GPa for the reservoir. Fig. 4a shows the simulation results for the surface uplift for both KB501 and KB503 using Young's modulus of 6 GPa, and Fig. 4b shows the simulation results using Young's modulus of 10 GPa. By using Young's modulus of 10 GPa for the reservoir, Sim-KB503-E10 and Sim-KB501-E10 match the measured surface uplift data quite well. The significant impact of Young's modulus on the simulation results of the surface uplift demonstrates the need for further investigation on the sensitivity of surface uplift to the other hydrogeological and geomechanical properties. Another field measurement was the maximum change in the pore pressure with respect to the initial pore pressure within the reservoir. That was reported to be approximately 10 MPa. Fig. 5 shows the maximum pore pressure within the reservoir throughout 3-year simulation period. The maximum change within the reservoir is approximately 9 MPa, which is in the range of the measured data from the field. Fig. 6 shows the change in the pore pressure with respect to the initial pore pressure, throughout the depth at year 3. Looking closely at Sim-KB503-E10, where the caprock and base permeabilities are 1.0×10−19 m2, the diffusion of pressure is symmetric with respect to the reservoir (between two red lines). However, in Sim-KB501-E10, where the caprock permeability is set to 1.0×10−21 m2, the diffusion of pore pressure is no longer symmetric on both sides of the reservoir. This is mainly due to the fact that the permeability of the base aquifer remained the same (1.0×10−19 m2). It should be noted that multiple levels of refinement were performed on the base mesh for KB503. The refinements were uniform in all directions. The results showed the base mesh (used above) is sufficient to be used for the rest of the study. Computational domain may artificially reduce the injection-induced over pressures in the simulation. Therefore, the domain size was expanded to 15 km in horizontal directions for KB503. Based on the

(Pamukcu et al., 2011; Shi et al., 2013). Therefore, it is essential to investigate the importance of the constraint on the pore pressure and its impact on the behavior of the overall system. 2.4. Parameter estimation The parameter estimation package PEST (Doherty, 2011; Doherty and Hunt, 2010) was used to perform model calibration. As described in Table 2, five groups are considered to evaluate the impact of hydrogeological (kz, α) and geomechanical properties (b, E) of reservoir and caprock on surface uplift and pore pressure increase due to CO2 injection. Because of limited available observed data, different parameterizations were tested to systematically evaluate non-uniqueness of the parameter estimation. The evaluation was performed based on various sets of parameters. Group I calibrates hydrogeological properties (kz, α) of reservoir. Group II calibrates both hydrogeological and geomechanical properties (kz, α, b and E) of reservoir. For both groups I and II geomechanical and hydrogeological properties of caprock are prescribed. Group III calibrates both hydrogeological and geomechanical properties (kz, α, b and E) of caprock using fixed reservoir properties and Group IV calibrates hydrogeological properties (kz,res , kz,cap , α res , αcap ) with prescribed geomechanical properties (b and E). Group V studies the impact of different combinations of hydrogeological and geomechanical properties for both reservoir and caprock. For KB501 and KB503, the observed data for model calibration are the maximum surface uplift over 3 years. Additionally, the maximum pore pressure increase of 10 MPa in the reservoir due to CO2 injection was included as the observed data (ΔPmax,obs = 10 MPa ). To evaluate the impact of the inclusion of ΔPmax,obs on the calibration process with the observed uplift data, comparison of model calibration for all five groups with and without ΔPmax,obs was performed. The total objective function (Φtotal ) to be minimized is the weighted sum of squared errors between the vector of observed (obs) and simulated (sim) values (m1) as T

Φtotal = Φuplift + Φpressure = (m1obs − m1sim ) Q (m1obs − m1sim )

(7)

where Φuplift and Φpressure are the objective functions from two different observation groups, the diagonal matrix Q represents the square of the weight, and a superscript T represents the transpose operation. A weight factor of 0.01 is used for the maximum pressure change, compared to a unit weight factor for all uplift data, to ensure that the objective function values from two different groups (Φuplift and Φpressure ) are of similar magnitute. For nonlinear models, a parameter upgrade vector is computed from the residual vector, and then solved iteratively using the GaussNewton method with the Levenberg-Marquardt (LM) parameter (Doherty and Hunt, 2010). For these problems, gradient-based local optimization methods are likely to find a set of parameters at a local 5

Journal of Petroleum Science and Engineering xx (xxxx) xxxx–xxxx

P. Newell et al.

Fig. 3. Details of the impact of CO2 injection on the surface uplift (the reservoir region has been defined with two red lines) (a) Total vertical stress (σzz ) (b) Change in the pore pressure (ΔP ) (c) Effective vertical stress (σzzeff ). (d) CO2 saturation (Sn) (e) Vertical displacement (dz).

3.2. Geomechanical and hydrogeological properties

fact that injection takes place in an anticlinal structure, 15 km is about the maximum domain extent without intersecting the gas-water contact. The change in pore pressure throughout the depth of the system at x=0, x=2.5 km and x=5.0 km are shown in Fig. 7 after three years of injection. The change of the pore pressure at x=2.5 km and x=5.0 km at year 3 is 34% and 74%, respectively. However, the maximum change within the reservoir at the symmetry line (x=0) is not significantly affected by the boundary effect. This shows that the boundary effect is less, closer to the well. Additionally, the overall boundary effect on the pore pressure is less than 1 MPa, when the boundary was extended to 15 km.

3.2.1. Biot's coefficient at KB501 In order to maximize the effect of the pore pressure on the solid skeleton, it is often assumed that Biot's coefficient of all the layers are equal to one. However, Biot's coefficient of sandstone is between 0.4 and 0.6 (Zoback, 2010). Biot's coefficient is expected to change with solid compressibility. Therefore, it is quite reasonable to assume different Biot's coefficients for different layers of both KB501 and KB503. Havens (2002) has shown that Biot's coefficient varies from 0.2 to 0.8 within only 55 m (2020–2075 m) depth of the Bakken formation. Even though the Biot's coefficient and the bulk modulus are correlated, in this study, Biot's coefficient was changed independently from the 6

Journal of Petroleum Science and Engineering xx (xxxx) xxxx–xxxx

P. Newell et al.

Fig. 6. Details of the change in the pore pressure with respect to the initial pore pressure (the reservoir region has been defined with two red lines).

Fig. 7. Comparison of the change in the pore pressure for domains of size 5 km and 15 km at year 3 at three locations x=0, 2.5 and 5 km.

Fig. 4. Comparison of the simulations of the ground surface uplift and the measured data by InSAR: (a) Young's modulus of the reservoir=6 GPa. (b) Young's modulus of the reservoir=10 GPa.

Fig. 8. The effect of the Biot's coefficient on the surface uplift in KB501. This case has all the properties of the base case (Sim-KB503-E10), except the Biot's coefficients of reservoir and caprock were set equal to 0.75 and 0.55, respectively.

Fig. 5. Maximum pore pressure with respect to the initial pore pressure within the reservoir throughout the 3-year simulation for two different simulations: Sim-KB503 is intended to capture the surface uplift of KB503 and Sim-KB501 is intended to capture the surface uplift of KB501.

Biot's coefficients capture the uplift of KB501 quite well. The initial slope of the curve matches the initial uplift better than the Sim-KB501. Precisely, the error reduces by 40% within 3 years after injection started. Moreover, the maximum change in the pore pressure remains within the observed range (approximately 9 MPa). This case shows change in Biot's coefficient could differ between KB501 and KB503.

bulk modulus within the given layer. As an example, Biot's coefficient of the caprock and reservoir were set equal to 0.75 and 0.55, respectively, and other parameters remained the same as KB503 (Sim-KB501-Biot). These values were chosen to assure the maximum change in the pore pressure within the reservoir does not exceed 10 MPa. Fig. 8 shows the results of the simulations and the impact of the new Biot's coefficients on the surface uplift. The new values of the

3.2.2. Reservoir's permeability at KB501 Rutqvist et al. (2010) have chosen an isotropic reservoir permeability to satisfy the maximum change in the pore pressure. However, 7

Journal of Petroleum Science and Engineering xx (xxxx) xxxx–xxxx

P. Newell et al.

Fig. 10. The effect of anisotropic permeability of the reservoir on the surface uplift in KB503. This case has all the properties of the base case (Sim-KB503-E10), except the permeability field for the reservoir, which is defined in Table 3.

Fig. 9. The effect of anisotropic permeability of the reservoir on the surface uplift in KB501. This case has all the properties of the base case (Sim-KB503-E10), except the permeability field for the reservoir, which is defined as: kx=ky=1.8×10−14 m2 and kz=0.08×10−14 m2. This permeability field implies the anisotropy ratio of 22.5.

comparing to Sim-KB503-E10 is −40% and −6% for Sim-KB503AnisoPerm-1 and Sim-KB503-AnisoPerm-2, respectively.

Iding and Ringrose (2009, 2010) confirmed the nature of pre-existing fractures within the reservoir at In Salah. The image logs of the core data show these fractures are mainly oriented in the near-vertical direction. To investigate the effect of anisotropy in the intrinsic permeability, we set kx = k y = 1.8 × 10−14 m2 and kz = 0.08 × 10−14 m2 . This gives an anisotropy ratio of 22.5. It is noted that these values are not optimized, but the optimal values for these parameters are investigated in the inverse modeling later. In selecting these values we had two considerations: (1) horizontal permeability remains close to 1.3 × 10−14 m2 , and (2) maximum change in the pore pressure with respect to the initial value within the reservoir remains approximately 10 MPa. Fig. 9 shows the anisotropic permeability can capture the surface uplift of the KB501 reasonably well. In comparing with KB501, the error reduces by 2% up to almost a year after injection started, but then increases to 7% at the end of the 3 years passed the injection. However, the purpose was to illustrates the possibility of anisotropic intrinsic permeability field as another option to capture the surface uplift of KB501, and this example served the purpose. Moreover, the maximum change of the permeability remains within the acceptable range (approximately 9 MPa) and it occurs within the reservoir. It should be noted that in this case, the other parameters remained the same as Sim-KB503.

3.2.4. Reservoir's permeability and Biot's coefficient at KB501 This case evaluates the combination of the geomechanical (Biot's coefficient) and hydrogeological (anisotropic permeability) properties of reservoir to investigate their impact on the surface uplift. In this case, Biot's coefficient of caprock and reservoir were set to 0.75 and 0.55, respectively. Additionally, kx=ky=1.0×10−14 m2 and −14 2 kz=1.3×10 m , which implies an anisotropy ratio of 0.77 was used for the reservoir in KB501. This example investigates the impact of the combination of geomechanical and hydrogeological properties being different than the properties listed in Table 1. Fig. 11 shows the result of this case. The new properties capture the measured uplift data from InSAR reasonably well. The fitting error in comparing to Sim-KB501E10 has reduced by 17%. The foregoing parameter studies illustrated that the InSAR data can be fit equally well by variation of different sets of hydogeological or geomecahnical properties or both. This demonstrates that more rigorous parameter estimation is required to evaluate the impact of different properties on the surface uplift. 3.3. Parameter estimation Four key parameters (kz, α, b, and E) for reservoir (res) and caprock

3.2.3. Reservoir's permeability at KB503 The possibility of anisotropic permeability field was also tested for KB503. Based on some preliminary tests, two different anisotropic permeability fields were tested as listed in Table 3. Horizontal permeabilities were selected to be in the range of 1.0×10−14 m21.0 × 10−14 m2 , to assure the maximum change in the pore pressure within the reservoir is within the acceptable range of the observed data. Fig. 10 shows the anisotropy ratio could play an important role in terms of the surface uplift. Although the surface uplift is almost identical for both isotropic (Sim-KB503-E10) and anisotropic (Sim-KB503-AnisoPerm-2) cases, due to the observed heterogeneity of the rock matrix (Rutqvist et al., 2010; Iding and Ringrose, 2009), the anisotropic permeability field seems more realistic in In Salah field. It should be noted that the change in the error in Table 3 Different permeability fields for reservoir at KB503. Case

kx

ky

kz

α

Sim-KB503-E10

1.3 × 10−14

1.3 × 10−14

1.3 × 10−14

1.0

Sim-KB503-AnisoPerm−1

1.033 × 10−14

1.033 × 10−14

1.927 × 10−14

0.54

Sim-KB503-AnisoPerm−2

1.3 × 10−14

1.3 × 10−14

2.5 × 10−14

0.52

Fig. 11. The effect of combination of the geomechanical (Biot's coefficient) and hydrogeological (anisotropic permeability) properties of reservoir on the surface uplift in KB501. This case has all the properties of the base case (Sim-KB503-E10), except the permeability field for the reservoir, which was defined as: kx=ky=1.0×10−14 m2, kz=1.3×10−14 m2 and the Biot's coefficient of caprock and reservoir, which were set to 0.75 and 0.55, respectively.

8

Journal of Petroleum Science and Engineering xx (xxxx) xxxx–xxxx

P. Newell et al.

(cap) layers were estimated against the observed surface uplift data at KB501 and KB503, and the maximum change in the pore pressure (approximately 10 MPa). As described in Table 2, five different parameter groups are chosen to evaluate the impact of hydrogeological and geomechanical properties as well as the inclusion of ΔPmax,obs on parameter estimation. Since multiple sets of parameters are expected to match the set of observed data equally likely, it is important to evaluate the significance of different sets of parameters on model calibration. In Tables 4 and 5, the maximum change in the pore pressure (ΔPmax ) was computed with calibrated parameter values. The final values of estimated parameters were selected when the value of the uplift objective function (Φuplift ) converged to less than 10−4, except in Group III for KB503, which had a bad convergence. The estimated parameters are presented as gray-shaded values in Tables 4 and 5. 3.3.1. Low uplift at KB501 Depending on the initial values and a set of estimated parameters, the kz value of the caprock (kz,cap ) ranged from 10-21 to 10-19 m2. The kz value of the reservoir (kz,res ) has a narrow range around a reference value of 1.3 × 10−14 m2 except in Group I. For all cases in Groups I-IV, the horizontal permeability of the reservoir (kz,res × α res ) are very similar to the reference value, while kz,cap values are low to match the low uplift at KB501. This indicates that lower kz,res values with ΔPmax constraint were compensated by higher α res values to dissipate the pore pressure increase from the injection zone to match the low uplift under the low caprock permeability during model calibration. However, KB501-4 (Group II) and cases in Group V show lower α res values and particularly, KB501-4 and KB501-10 with ΔPmax constraint have lower Biot's coefficients. As discussed in the previous section, the Biot's coefficient and pore pressure increase are inter-related through the effective stress. The lower Biot's coefficient and higher vertical permeability have an opposite effect on the maximum uplift. KB501-10 clearly shows that the positive impact of higher kz,cap and lower α res on uplift was reduced by the negative impact of lower bcap value on uplift, matching the low uplift very well (Table 4 and Fig. 12). This is also supported by the negative correlation coefficient (−0.90) between kz,cap and bcap . In addition, higher variance values for kz,cap (43.5) and αcap (3.95) than those for kz,res (5.03) and α res (0.84) indicates that caprock hydrogeological parameters are less constrained to the observed data than reservoir properties. All cases for KB501 match the surface uplift relatively well both in

Fig. 12. Comparison of observed and modeled uplifts for KB501 (symbol only) and KB503 (symbol+line) with different calibrated sets of parameters. Estimated parameters are presented in Table 4 and Table 5.

terms of Φuplift (Table 4) and through visual comparison (Fig. 12), confirming that multiple sets of parameters can match the observed data well. However, comparison of model calibration with and without ΔPmax constraint shows that computed ΔPmax values using calibrated parameters without the constraint are much lower than those with the constraint. For example, Groups II (cases 4 and 5) and V (cases 10 and 12) show that with the pore pressure constraint both Biot's coefficient and anisotropy ratio parameters are calibrated differently, reconfirming that the inclusion of the pore pressure data is critically important to constrain the parameter solution direction during model calibration. As demonstrated in Fig. 11, both geomechanical (bres ) and hydrogeological (α res ) parameters play key roles in the pore pressure propagation through reservoir and the magnitude of the surface uplift. Therefore, parameter estimation results highlight the significance of coupled impact of hydrogeological and geomechanical properties during GCS. With ΔPmax constraint Young's modulus are very close to the base value for all layers, indicating the base values of Young's modulus may represent the geomechanical properties well.

3.3.2. High uplift at KB503 In contrast to the KB501 cases, the surface uplift in KB503 was

Table 4 Calibration results for KB-501 cases. Group

Group I

Case

1

Group II 2

Group III

Group IV

Group V

3

4

5

6

7

8

9

10

11

12

Calibrated Model Parameter Values

kz,cap (m2)

1.00E−20

1.00E−21

1.00E−20

1.00E−20

1.00E−20

5.07E−21

2.29E−21

2.38E−21

4.47E−20

9.15E−20

1.06E−19

8.12E−20

kz,res (m2) αcap

3.34E−15 1.000 4.274 1.000

1.68E−15 1.000 8.543 1.000

1.61E−14 1.0000 1.036 1.000

2.39E−14 1.0000 0.429 1.000

1.14E−14 1.0000 1.158 1.000

1.75E−14 0.458 1.000 1.000

1.30E−14 0.516 1.000 1.000

1.75E−14 0.869 1.000 1.000

4.27E−14 1.452 0.547 1.000

1.59E−14 0.972 0.67 0.644

2.06E−14 1.219 0.481 0.6

1.51E−14 1.024 1.095 0.681

1.000 bres 20 Ecap (GPa) E res (GPa) 10 Computed Resultsa

1.000 20

1.000 20

0.525 20

1.000 20

0.55 17.8

0.550 17.8

1.000 20

1.000 20

0.550 20

0.550 20

1.000 20

10

10

9.76

9.84

10

10

10

10

10

10

10

ΔPmax b Φ uplift

9.18 7.32E−05

9.44 6.97E−05

6.94 6.43E−05

9.94 5.44E−05

8.8 9.03E−05

6.58 9.12E−05

8.8 5.41E−05

6.58 7.15E−05

4.54 7.10E−05

9.99 4.54E−05

10.3 4.51E−05

7 6.00E−05

Φ pressure Φ total

6.77E−05

3.16E−05

n.a.

1.90E−05

n.a.

1.17E−03

1.45E−05

1.17E−03

n.a.

1.40E−08

n.a.

n.a.

1.41E−05

1.01E−04

6.43E−05

7.34E−05

9.03E−05

1.26E−03

1.99E−04

1.24E−03

7.10E−05

4.54E−05

4.51E−05

6.00E−05

αres bcap

Highlighted parameters in gray are estimated with other parameters fixed. Cases without pressure constraint show no objective function value as indicated by n.a.. a Final estimated values are the same as the initial values of parameters. b ΔPmax is the computed maximum pore pressure change (in MPa) with calibrated parameter values.

9

Journal of Petroleum Science and Engineering xx (xxxx) xxxx–xxxx

P. Newell et al.

Table 5 Calibration results for KB-503 cases. Group

Group I

Case

1

Group II 2

Group III

Group IV

Group V

3

4

5

6

7

8

9

10

11

12

Calibrated Model Parameter Values

kz,cap (m2)

1.00E−19

1.00E−19

1.00E−19

1.00E−19

1.00E−19

4.57E−20

1.45E−19

1.72E−19

2.77E−20

1.29E−19

1.20E−19

1.39E−19

kz,res (m2) αcap α res bcap

2.03E−14 1.000 0.511 1.000

1.93E−14 1.000 0.536 1.000

1.08E−14 1.000 0.895 1.000

4.09E−14 1.000 0.225 1.000

1.33E−14 1.000 0.623 1.000

1.30E−14 0.844 1.000 0.986

3.16E−14 0.750 1.000 0.750

1.02E−14 0.311 1.079 1.000

8.68E−15 0.250 0.872 1.000

2.26E−14 1.015 0.431 1.000

1.58E−14 1.485 0.644 1.000

1.97E−14 1.715 0.283 0.862

1.000 20

1.000 20

0.879 20

0.779 20

0.550 17.8

0.550 18.6

1.000 20

1.000 20

0.550 17.8

0.550 17.8

0.550 20.9

10

10

9.83

10.85

10

10

10

10

10

10

10

10.10 9.92E−05

11.30 9.47E−05

10.03 8.96E−05

12.50 8.25E−05

8.75 3.54E−05

6.57 1.80E−05

10.31 9.79E−05

14.08 9.68E−05

10.3 7.98E−05

10.4 9.29E−05

−04

−03

−06

−05

−05

1.000 bres 20 Ecap (GPa) E res (GPa) 10 Computed Results

ΔPmax a Φ uplift Φ pressure Φ total

10.02 9.93E−05 4.00E

−08

17.10E−05

−08

n.a

n.a.

9.00E

9.92E−05

9.47E−05

8.96E−05

n.a.

1.55E

1.18E

8.25E−05

5.09E−04

1.36E−03

9.61E

9.03E−05

n.a.

1.09E

9.68E−05

9.07E−05

1.85E

1.11E−05

16.2 7.10E−05 n.a. 7.10E−05

Highlighted parameters in gray are estimated with other parameters fixed. Cases without pressure constraint show no objective function value as indicated by n.a.. a ΔPmax is the computed maximum pore pressure change (in MPa) with calibrated parameter values.

been applied for characterizing hydrogeological features around other injection wells (Shi et al., 2012, 2013). Based on cross-validation and analysis of variance methods with 13 input parameters for coupled multiphase flow and geomechanical simulations with varying injection rate in a 2D layered system with a 200 m of reservoir layer and a 100 m of caprock layer, Bao et al. (2013) concluded the three most significant parameters to surface uplift and pore pressure increase are reservoir porosity, permeability, and injection rate. Caprock properties are generally less significant than reservoir properties. In this work, however, caprock properties are very important to match the surface uplift with pore pressure constraint. This difference can be attributed to the thickness of reservoir and caprock, which results in different sensitivity of parameters to surface uplift and pore pressure distribution. As expected for strong nonlinear problems, model calibration was significantly influenced by the starting values. For example, the high surface uplift coupled with the maximum pore pressure constraint results in very narrow solution space and lower caprock permeability as a starting point results in poor convergence. Overall, calibration results with different starting parameters and different calibration groups provide a range of estimated parameters matching both observed data. Recently, Ringrose et al. (2013) summarized the key elements of site characterizations and monitoring technologies applied at In Salah site, highlighting the significance of integrated approach to the development of monitoring, modeling and verification (MMV) for CO2 storage. As suggested by Ringrose et al. (2013), this work also supports that calibration of coupled reservoir and geomechanics model integrated with geophysical characterization techniques will provide a practically efficient tool for developing actual operations of CO2 injection and monitoring/evaluating its associated risk assessments.

high, estimating high kz,cap (all cases except case 11), high Biot's coefficient (Groups II and V), and/or low Young's modulus (cases 10 and 11). As in the KB501 cases, all cases match the surface uplift well (Table 5 and Fig. 12). For all starting values and different sets of parameters, kz,res has a very narrow range of estimated values close to the reference value (1.3 × 10−14 m2 ). As shown in Fig. 5, the pore pressure corresponding to the high surface uplift at KB503 propagates significantly in the caprock layer. This indicates reservoir and caprock properties at KB503 might be more inter-related. For example, compared to correlations between hydrogeological and geomechanical parameters for the KB501-10 (e.g., −0.2 for α res -bcap and 0.24 for kz,res -bcap ), the KB503-10 case has higher correlations (e.g., −0.89 for α res -bcap and 0.91 for kz,res -bcap ). The importance of multiple observed data is also shown by computed ΔPmax values. Without the pore pressure constraint, computed ΔPmax values were higher for all cases in KB503 than in KB501. It should be noted that the surface uplift tends to calibrate the model without direct information regarding the pressure propagation or CO2 plume development. This strongly suggests that calibration of both hydrogeological and geomechanical properties together with multiple datasets can constrain parameter space closer to actual physical setting.

3.3.3. Implications for practical applications Pamukcu et al. (2011) calibrated the dual porosity and permeability model by trail-and-error against the bottom hole pressure and the CO2 breakthrough time. The model parameters consist of permeabilities of reservoir and overlying lower caprock layers. Their manual sensitivity results show that the reservoir has a higher anisotropy ratio (i.e., higher horizontal permeability) from 10 to 100 for matrix and fracture, respectively. In this work, it was found that high α res values can be obtained with reservoir hydrogeological parameters (Group I in KB501) and a lower E res of 6 GPa (results not shown), which is lower than actual value of 10 GPa. The high α res reported in Pamukcu et al. (2011) may result from an incomplete set of model parameters and/or an improper value of geomechanical parameters. In contrast, our results indicate that with a proper E res value (approximately 10 GPa) automatic calibration results with the surface uplift data and pore pressure data may reflect an average value of matrix and fracture permeability. At In Salah, the site characterization based on one injection well has

4. Concluding remarks In this work, the influence of geomechanical and hydrological properties on surface uplift at In Salah was numerically investigated to identify the impact of key geomechanical and hydrological parameters such as Biot's coefficient and permeability anisotropy ratio under various conditions. The maximum change in the pore pressure due to CO2 injection was used to constrain the inverse model for both low and high surface uplift data. Forward simulation results with representative parameter values in the literature matched surface 10

Journal of Petroleum Science and Engineering xx (xxxx) xxxx–xxxx

P. Newell et al.

carbon sequestration. Adv. Water Resour. 54 (0), 133–148. Genuchten, M.T.V., 1980. A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Sci. Soc. Am. J. 44 (5), 892–898. Havens, J., 2012. Mechanical properties of the bakken formation. Master’s thesis, Colorado school of Mines. Iding, M., Ringrose, P., 2009. Evaluating the impact of fractures on the long-term performance of the In Salah CO2 storage site. Energy Procedia 1 (1), 2021–2028. Iding, M., Ringrose, P., 2010. Evaluating the impact of fractures on the performance of the In Salah CO2 storage site. Int. J. Greenh. Gas. Control 4 (2), 242–248. Khakim, M.Y.N., Tsuji, T., Matsuoka, T., 2012. Geomechanical modeling for insarderived surface deformation at steam-injection oil sand fields. J. Pet. Sci. Eng. 96– 97, 152–161. Martinez, M.J., Stone, M.C., Notz, P.K., Tuner, D.Z., Hopkins, P.L., Subia, S., Jove-Colon, C., Moffat, H.K., Bean, J.E., Dewers, T., Klise, K., Red-Horse, J., R, R.F., Mesh, M., Davison, S., Yoon, H., Carnes, B., Alger, N., Bishop, J.E., Newell, P., 2011. Computational thermal, chemical, fluid and solid mechanics for geosystems management. SAND2011-6643. Martinez, M.J., Newell, P., Bishop, J., Turner, D., 2013. Coupled multiphase flow and geomechanics model for analysis of joint reactivation during CO2 sequestration operations. Int. J. Greenh. Gas. Control 17, 148–160. McKenna, S.A., Pike, D.Q., 2013. Fluid pressure redistribution events within a fault: impact of material property correlation. In: Mishra, P.K., Kuhlman, K.L. (Eds.), Advances in Hydrogeology. Springer, New York, 57–72. Nanayakkara, A.S., Wong, R.K., 2009. How far does surface heave propagate? A discussion on analytical and numerical modeling of the surface heave induced by subsurface fluid injection. KSCE J. Civ. Eng. 13 (4), 297–303. Notz, P., Subia, S., Hopkins, M., Moffat, H., Noble, D., 2007. Aria 1.5: User manual. SAND2007-2734. Pamukcu, Y., Hurter, S., Jammes, L., Vu-Hoang, D., Pekot, L., 2011. Characterizing and predicting short term performance for the In Salah krechba field CCS joint industry project. Energy Procedia 4 (0), 3371–3378, (10th International Conference on Greenhouse Gas Control Technologies). Preisig, M., Prevost, J.H., 2011. Coupled multi-phase thermo-poromechanical effects. case study: CO2 injection at In Salah, Algeria. Int. J. Greenh. Gas. Control 5 (4), 1055–1064. Raikes, S., Mathieson, A., Roberts, D., Ringrose, P., 2008. Integration of 3d seismic with satellite imagery at in salah CO2 sequestration project, Algeria. Gulf Atomics ReportGAC19579 Prepared for the Defense Nuclear Agency under contract DNA001086C-0277. Ringrose, P., Mathieson, A., Wright, I., Selama, F., Hansen, O., Bissell, R., Saoula, N., Midgley, J., 2013. The in salah CO2 storage project: Lessons learned and knowledge transfer. Energy Procedia 37 (0), 6226–6236, (11th International Conference on Greenhouse Gas Control Technologies). Rutqvist, J., Vasco, D.W., Myer, L., 2009. Coupled reservoir-geomechanical analysis of CO2 injection at InSalah, Algeria. Energy Procedia 1 (1), 1847–1854. Rutqvist, J., Vasco, D.W., Myer, L., 2010. Coupled reservoir-geomechanical analysis of CO2 injection and ground deformations at In Salah, Algeria. Int. J. Greenh. Gas. Control 4 (2), 225–230. Shi, J., Sinayuc, C., Durucan, S., Korre, A., 2012. Assessment of carbon dioxide plume behaviour within the storage reservoir and the lower caprock around the kb-502 injection well at in salah. Int. J. Greenh. Gas. Control 7 (0), 115–126. Shi, J., Smith, J., Durucan, S., Korre, A., 2013. A coupled reservoir simulationgeomechanical modelling study of the CO2 injection-induced ground surface uplift observed at krechba, In Salah. Energy Procedia 37 (0), 3719–3726. Smith, J., Durucan, S., Korre, A., Shi, J., Sinayuc, C., 2011. Assessment of fracture connectivity and potential for CO2 migration through the reservoir and lower caprock at the In Salah storage site. Energy Procedia 4 (0), 5299–5305, (10th International Conference on Greenhouse Gas Control Technologies). Smith, J., Durucan, S., Korre, A., Shi, J., 2013. A methodology to assess increased storage capacity provided by fracture networks at CO2 storage sites: application to In Salah storage site. Energy Procedia 37 (0), 3711–3718. Tavakoli, R., Yoon, H., Delshad, M., ElSheikh, A.H., Wheeler, M.F., Arnold, B.W., 2013. Comparison of ensemble filtering algorithms and null-space monte carlo for parameter estimation and uncertainty quantification using CO2 sequestration data. Water Resour. Res. 49 (12), 8108–8127. Vasco, D.W., Novali, A.F., 2008. Reservoir monitoring and characterization using satellite geodetic data: interferometric synthetic aperture radar observations from the krechba field, Algeria. Geophysics 73 (6), WA113–WA122. Vasco, D., Ferretti, A., Novali, F., 2008. Estimating permeability from quasi-static deformation: temporal variations and arrival time inversion. Geophysics 73 (6), O37–O52. Vasco, D.W., Rucci, A., Ferretti, A., Novali, F., Bissell, R.C., Ringrose, P.S., Mathieson, A.S., Wright, I.W., 2010. Satellite-based measurements of surface deformation reveal fluid flow associated with the geological storage of carbon dioxide. Geophys. Res. Lett. 37 (3). Verdon, J.P., Kendall, J.M., Stork, A.L., Chadwick, R.A., White, D.J., Bissell, R.C., 2013. Comparison of geomechanical deformation induced by megatonne-scale CO2 storage at sleipner, weyburn, and in salah (supplement). Proceeding of National Academy of Science of United States of America, Bremerhaven. Wainwright, H.M., Finsterle, S., Zhou, Q., Birkholzer, J.T., 2013. Modeling the performance of large-scale CO2 storage systems: a comparison of different sensitivity analysis methods. Int. J. Greenh. Gas. Control 17 (0), 189–205. Yoon, H., McKenna, S.A., 2012. Highly parameterized inverse estimation of hydraulic

uplifts reasonably well maintaining the maximum change in the pore pressure within the observed field data. In particular, the simulation results matched the surface uplift better, when different Biot's coefficients for reservoir and caprock layers were used. This result highlights the importance of Biot's coefficient in coupled reservoir and geomechanical models of subsurface system including CCS units. For inverse modeling multiple sets of parameters were able to match the observed data equally well. The vertical intrinsic permeability and Young's modulus of the reservoir remained close to 10– 14 mD and 10 GPa, respectively for a majority of cases. In particular, the inclusion of the pore pressure data was critically important to constrain the parameter solution within physically reasonable ranges, suggesting the inclusion of the pore pressure constraints is essential to estimate the proper values of coupled flow and geomechanical properties associated with different surface uplift data. The results also indicated the robustness of these parameters over many variations of other parameters in order to satisfy the pore pressure constraint. Additionally, some of the cases in parameter estimation of reservoir permeability suggested consistency with the fracturing proposed by Ringrose et al. (2013). Although the numerical domain was a simplified layered system and the local heterogeneity and fracture systems are not included explicitly, this study suggests that given limited data such as point measurement of deformation and bottom hole pressure at the injection well, proper consideration of model parameters such as Biot's coefficient can enhance parameter estimation of the geomechanical response during GCS. This work also implies that parameter estimation with multi-objective functions such as surface uplift and pore pressure data can be utilized as in Pareto analysis to determine optimization under uncertainty for operational conditions such as CO2 injection rates and/or failure criteria . Acknowledgements This work is supported as part of the Center for Frontiers of Subsurface Energy Security, an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences under Award Number DE-SC0001114. Sandia National Laboratories is a multi-program laboratory managed and operated by Sandia Corporation, a wholly owned subsidiary of Lockheed Martin Corporation, for the U.S. Department of Energy's National Nuclear Security Administration under contract DE-AC0494AL85000. References Aoyagia, R., Imai, R., Rutqvist, J., Kobayashi, H., Kitamura, O., Goto, N., 2013. Development of TOUGH-FrontISTR, a numerical simulator for environmental impact assessment of CO2 geological storage. Energy Procedia 37, 3655–3662, (11th International Conference on Greenhouse Gas Control Technologies). Bao, J., Hou, Z., Fang, Y., Ren, H., Lin, G., 2013. Uncertainty quantification for evaluating impacts of caprock and reservoir properties on pressure buildup and ground surface displacement during geological CO2 sequestration. Greenh. Gases: Sci. Technol. 3 (5), 338–358. Cavanagh, A., Ringrose, P., 2010. In salah high resolution heterogeneous simulations of co2 storage. Search and discovery article #80092. Deflandre, J., Estublier, A., Baroni, A., Daniel, J., Adjémian, F., 2011. In Salah CO2 injection modeling: a preliminary approach to predict short term reservoir behavior. Energy Procedia 4 (0), 3574–3581, (10th International Conference on Greenhouse Gas Control Technologies). Doherty, J., Hunt, R.J., 2010. Approaches to highly parameterized inversion: A guide to using pest for groundwater-model calibration. USGS Scientific Investigations Report 2010–5169. Doherty, J., 2011. PEST: Model independent parameter estimation. Eiken, O., Ringrose, P., Hermanrud, C., Nazarian, B., Torp, T.A., Høier, L., 2011. Lessons learned from 14 years of CCS operations: Sleipner, in salah and snøhvit. Energy Procedia 4 (0), 5541–5548, (10th International Conference on Greenhouse Gas Control Technologies). Espinet, A.J., Shoemaker, C.A., 2013. Comparison of optimization algorithms for parameter estimation of multi-phase flow models with application to geological

11

Journal of Petroleum Science and Engineering xx (xxxx) xxxx–xxxx

P. Newell et al.

536–553. Zhou, X., Burbey, T.J., 2014. Distinguishing fluid injection induced ground deformation caused by fracture pressurization from porous medium pressurization. J. Pet. Sci. Eng. 121, 174–179. Zoback, M.D., 2010. Reservoir Geomechanics. Cambridge University Press.

conductivity and porosity in a three-dimensional, heterogeneous transport experiment. Water Resour. Res. 48 (10), (W51036:1-17). Yoon, H., Hart, D.B., McKenna, S.A., 2013. Parameter estimation and predictive uncertainty in stochastic inverse modeling of groundwater flow: comparing nullspace monte carlo and multiple starting point methods. Water Resour. Res. 49 (1),

12